MiniMax M1 — reviews, specs & pricing
MiniMax's open-weight reasoning model with extremely long context.
Summary
MiniMax M1 is a 456B parameter (45.9B active) hybrid-attention MoE reasoning model supporting up to 1M tokens of context, trained with efficient large-scale reinforcement learning. It's released under Apache 2.0. It targets long-context agentic and reasoning tasks.
Sample use case
Used for long-document analysis, extended agentic workflows, and research on efficient long-context RL training. Suited for tasks needing million-token context.
Specifications
- Provider: minimax
- License: open
- Parameters: 456B
- Context: 1000k tokens
- Input price: $0.4/M tok
- Output price: $2.2/M tok
- Released: 2025-06-16
Pros
- 1M token context
- Apache 2.0 license
- Efficient RL training approach
Cons
- Requires heavy compute to self-host
- Newer, fewer third-party evaluations
- Ecosystem still developing
Average rating 0.0 from 0 community reviews on Reviuws.